Using: 'Chi square' & 'ANOVA' tests


http://members.tripod.com/~wu_wei2/Chapter_7_arguments.html

One can disprove this sort of statistical analysis by performing their own control. Flip a coin 10 times. Record how many heads and tails you receive. By your logic, one should always receive 5 heads and 5 tails, since the probability is 50% for each. Try doing this test 10 times. How often do you get 5 of each?

Better yet, get a six-sided die. Roll it 12 times, you should get two of each result. Try that 10 times. How often do you get 2 of each roll?

Or draw a random card from a deck. How often do you get a heart, a king, the king of hearts, or a joker?

Probabilities are the likelihood, not a guarantee of what the results will be. Probabilities are usually only accurate over EXTREMELY large data sets. There will be many trends that seem to come and go at every point in between...

AE,

Ah, my young friend, you are forgetting about levels of statistical significance.... you know .05,.01,.001....

[quote]Probabilities are the likelihood, not a guarantee of what the results will be. Probabilities are usually only accurate over EXTREMELY large data sets. There will be many trends that seem to come and go at every point in between...[/quote]

Inherent within statistcal analysis are the procudures for seperating 'chance' from 'design' or 'influence, or 'cause'.
It's a measure of how strong the influence of variable is in terms of the experimenter receiving his data from the experimental design.

Its all about levels of statitical significance.

Next?

[quote]Number recurrences resulting from random numbers generated in a range 1 - 64 will exhibit the same number of recurrences as were seen in actual oracles generated between April 1990 and November 1997[/quote]

I was arguing this first basic contention...

Ae,

Why would you argue that... 1 to 64 simply becomes the test size of your variables influence. The slice size, if you will, of the whole pie (infinity) that you are measuring. That's a legitimate experiemntal design.

If you measure the influence of a slice (1-64) of infinity, and the effect is real in the slice then you must accept that perhaps (ie confirmed with further study)the effect also is occuring in infinity...in approximately the same way.

n'est-ce pas?

You apparently have poor reading comprehension skills.

[quote]Number recurrences resulting from random numbers generated in a range 1 - 64 will exhibit the same number of recurrences as were seen in actual oracles generated between April 1990 and November 1997[/quote]

^This says that the number of oracle drawings of a particular type have to match the number that SHOULD be drawn based on the probability of drawing them. I explained how presumptuous this was and gave examples of how one could easily disprove this assumption. It has nothing to do with having 64 options. You could have 2 or 10,000,000. Probability only indicates chance of occurrence...

Ae,

As I read it, there is no difference.

What he is saying is - he did a whole bunch of individual I Ching 'casting the coin' trials trials between 1990-97. And then he randomly generated a whole bunch of trials (with numbers 1 - 64) to compare the incidences of 1-64 numbers. He is just establishing the comparability of his sample with the infinite... justifying, if you will, his null hypotheis, thats all.

But yes, I do have poor reading comprehension skills ...hating reading until the internet came along.